Fetching the paper…

GLM: General Language Model Pretraining with Autoregressive Blank Infilling · Around